Effective Development of Fuzzy-Logic Rules for Real-time Control of Autonomous Vehicles

Abstract The design of the fuzzy rules for fuzzy-logic controllers influences the ultimate performance of real-time control systems. In general, different sets of fuzzy rules are formed according to the designers' heuristic thinking, knowledge, or experience about the system. It would be difficult for others to modify the rules so formed unless the process of capturing the original knowledge and experience can be systematically described. This paper proposes a direct and effective method to obtain an initial set of fuzzy rules from expert knowledge for designing linear controllers of autonomous vehicles, and to refine the rules to yield high-performance nonlinear controllers. The design considerations and performance achieved are discussed. The fuzzy-logic controller is able to cope effectively with uncertainties in practical situations.